Vision, Learning and Graphics group,
Dept. of Computer Science,
Courant Institute of Mathematical Sciences,
New York University
Facebook AI Lab
Director of Graduate Studies for
Master of Science in Data Science
I'm joining Facebook!
I'm happy to announce that I am joining Facebook's new AI Group, a research laboratory with the long term goal of making major advances in the field. I'll be working alongside Yann LeCun, who will be leading the Group (Link). Facebook first announced the AI Group in September to work on problems in deep learning, machine learning and computer vision.
I will be working part time at Facebook until May, when my sabbatical starts and then I will be based full time at Facebook's new office at Astor Place, one block away from NYU. After my sabbatical, I will continue to work both at Facebook and NYU.
Facebook is building the AI Group here in New York, in Menlo Park and in London. This is an exciting time for the field, and I'm looking forward to getting started.
My research is in the areas of Machine Learning and Computer Vision. I am particularly interested in applying Deep Learning methods to object recognition. I also work on low-level vision problems, with applications to computational photography and astronomy.
Deep Learning for Computer Vision
NIPS 2013 Tutorial [Slides]
Online Recognition Demo
See our deep convolutional network demo here. This network achieves 16.5% top-5 error on the Imagenet 2012 classification challenge, around 2% better than the network of Krizhevsky et al. (NIPS 2012).
International Conference on Learning Representations 2014
I am one of the Program Chairs for a new conference on feature learning and representation learning. The submission deadline is December 20th. Check out the website.
Pre-prints of recent research can be found on arXiv: Link
Visualizing and Understanding Convolutional Networks
Matt Zeiler and Rob Fergus,
arXiv pre-print, Nov 2013, PDF
Reconnaissance of the HR 8799 Exosolar System I: Near IR Spectroscopy
Indoor Segmentation and Support Inference from RGBD Images
Adaptive Deconvolutional Networks for Mid and High Level Feature Learning
Learning Invarance through Imitation
Blind Deconvolution using a Normalized Sparsity Measure
Dark Flash Photography
80 million tiny images: a large dataset for non-parametric object and scene recognition